Skala community edition update: GPU4PySCF support is in ✅
That means Skala can plug into GPU-accelerated PySCF DFT workflows via GPU4PySCF (API-compatible with PySCF).
Install docs are in the repo README: github.com/microsoft/sk...
#DFT #PySCF #GPU #GPU4PySCF #CompChem
This week I will be in Lausanne for Psi-K. Any bluesky tweeps around? 😄
Come join our wonderful team!
Interested in our mission to make DFT more accurate and push what’s possible in quantum chemistry? Do you want to directly contribute? We're hiring a senior software engineer and a senior researcher:
jobs.careers.microsoft.com/global/en/jo...
jobs.careers.microsoft.com/global/en/jo...
🚀 After two+ years of intense research, we’re thrilled to introduce Skala — a scalable deep learning density functional that hits chemical accuracy on atomization energies and matches hybrid-level accuracy on main group chemistry — all at the cost of semi-local DFT ⚛️🔥🧪🧬
Look Around You is still one of my favourite comedy bits.
Watch the pilot on Calcium and see if we share the same humor: youtu.be/FBaVwwuErmU
Just received word that Boltzmann is still trapped in the coffee corner.
Alles Gute zum Geburtstag! At least if my German is not horribly off.
I'm curious to know if anybody has done a comparison of the performance of torch.autograd and torch.func regarding speed, memory usage and so on. I like torch.func a lot, but it won't be so easy to convince others without showing (at least) equivalence.
Beautiful, this asks for a "don't talk to me or my son ever again" meme.
I should buy that book. 🤔
Yes, this I absolutely agree with. The data isn't there to learn this. In a way this experiment is weird in that one usually tries to generalize from less to more atoms and not the other way round.
It has a more complicated PES in the equilibrium region, but I wouldn't call that less smooth, I think. A simple MLP of the internuclear distance would easily fit it in any case. The question is if that's the case for the MLIP models as well, though.
Cr2 is hard from the quantum chemistry perspective, but is it the same from a force field perspective? Sure there is this weird decay, but a function of R is a function of R.
I guess the issue here is that if you train on MD data, which I think is usually the case (?), one simply doesn't see these distances. Perhaps systematically exploring also high energy regions in the PES could alleviate this?
"Overgangsfase"? Dit was in the good old days sowieso veel normaler, met één app posten op alle sociale media.
Waarvoor dank.
Too dark? I don't know what the level of darkness tolerated on bluesky is yet.
It's also similar to children in the sense that it's hard to erase if you made a mistake.
Right, but NVIDIA essentially provides only a C++ API, so it still matters how well that was tied into the other language, no?
From what I saw it's not great for GPU programming, am I correct in that?
Okay, that's fine, but only if it's in a 6-31G* basis and not an orbital more.
I've tried to get this point across multiple times, but usually failed at the "but CCSD(T) has an underlying wave function" argument. If only I could have those arguments again!
In this post-truth era we simply decide it to be correct. Repeat after me: CCSD(T) is the gold standard of quantum chemistry, CCSD(T) is the gold standard of quantum chemistry, CCSD(T) is the gold standard of quantum chemistry.
Frederick's adorable!
